DTW Based Automatic Passenger Counting Method Research Using Infrared Sensors

The development of Intelligent Transportation System (ITS) has accelerated the steps of transport system to enter the information era. As one of the most important parts of traffic information, passenger-counting system is an unsolvable urgent project, which is regarded important by related departments. There are many flaws in traditional counting methods. So a cheap, real-time, feasible method should be researched. A counting method based on infrared sensors would be one of the best choices because of its strongpoint.Dynamic time warping (DTW) is widely used in sound signal processing field, and it achieves a high accuracy in isolated word recognition. Passenger-passing signal is comparable with sound signal, so DTW is introduced into automatic passenger counting (APC) system. Clustering algorithm is used to choose adaptive signal as references. And sample signal is endpoint detected to pick up valid signal, which is used to match references to generate result. DTW strongly depends on point-detection. So it is necessary to detect endpoint well to achieve a higher accuracy. A new point-detection method is brought forward based on endpoint features research. Local search scope of DTW is improved to make this algorithm suit this field well.Both active and passive sensors are used in this project. Endpoint detection and DTW are realized in VC environment. Off-line simulation is designed to observe the validity vividly. The result shows that it could detect endpoint and count well and the accuracy reaches 97%. Counting method using infrared sensors has many advantages, such as low cost, safety and non-contact. It can also be installed in other public places like stores, stadiums and so on. Experts predict that it has broad market prospect.